Workspace instructions vs. custom instructions.

LADLE has two places where you can tell the assistant how to behave:

- **Custom instructions** (Settings → Behavior) are yours. They apply to every chat you send, in every workspace and in Personal. - **Workspace instructions** (Workspace settings → Workspace instructions, admin only) apply to every workspace member's chats sent inside that workspace. They do not apply to members' Personal chats or their chats in other workspaces.

Both are text blocks up to 8,000 characters. The assistant reads them in a specific order:

1. Base persona (LADLE's own rules — honesty, safety, formatting). 2. Response language and style directives. 3. Your memory. 4. Your custom instructions. 5. Workspace instructions (if inside a workspace). 6. Project instructions (if inside a project). 7. Workspace knowledge (if inside a workspace, and workspace has notes). 8. Project knowledge (if inside a project, and project has files).

Because your custom instructions come before workspace instructions, personal preferences win conflicts. Say your custom instructions say "reply in bullet points" and the workspace instructions say "reply in prose" — the assistant will use bullet points for your chats.

**When to put a rule in workspace instructions instead of asking each member to add it to their own:** anything that's a shared truth (a glossary term, a product's canonical name, the tone the team commits to publicly), any policy that shouldn't drift member-by-member (must-cite requirements), any voice guide you want the AI to apply consistently across the whole team's work.

**When to keep a rule personal instead of workspace-level:** ergonomic preferences (bullet vs. prose, verbosity), individual workflow markers ("start with the answer, not the analysis"), personality preferences (formal vs. warm). These are rightly personal.

Editing workspace instructions is admin-only. Every workspace member sees the effect in their own chats without any action on their part.